Knowledge Engineering-Based Quality Evaluation of NRG Oncology RTOG 0522 Treatment Plans

Abstract

To evaluate radiotherapy treatment plan quality of IMRT plans submitted for RTOG 0522 clinical trial (A Randomized Phase III Trial of Concurrent Accelerated Radiation and Cisplatin versus Concurrent Accelerated Radiation, Cisplatin, and Cetuximab (C225) [Followed by Surgery for Selected Patients] for Stage III and IV Head and Neck Carcinomas). To use knowledge engineering based models from RapidPlan (built from HN002) for the quality study, and to re-optimize the plans by the model with the aim of further sparing the organs at risk and improve the quality of these plans in general. A head and neck RapidPlan model was initially built in Varian Eclipse treatment planning system (Version 13.6.15) using treatment plans from the NRGHN002 clinical trial (A Randomized Phase II Trial for patients with p16 positive, non-smoking-associated, locoregionally advanced oropharyngeal cancer). The model was duplicated and tuned to be used for QA and re-optimizing treatment plans from RTOG 0522. The modified model was then used to QA and re-optimize plans with deviation unacceptable scores (i.e., failed to pass the dosimetric compliance criteria of RTOG 0522). Initial quality review of 748 IMRT plans submitted to RTOG0522 show that 21 plans did not meet the spinal cord dose constraint (D0.03 cc <= 48 Gy), and 323 plans did not meet the dose constraint for parotid glands (one of the gland has mean dose less than 26 Gy). All the 21 plans that did not meet spinal cord constraint were pick for further analysis using the HN002 model, the results show that 15 of them has submitted plan dose volume histogram (DVH) fell above the predicted band. All the 21 plans were further optimized using model generated line objectives for organ at risks (OARs) (spinal cord, parotid glans, larynx, and NonPTV). All of plans achieved better scores after re-optimization (per-protocol or deviation acceptable). Average spinal cord Dmax [Gy] was decreased from 48.9 ± 2 Gy to 41.4 ± 4.8 Gy. Average parotids gland Dmean [Gy] were decreased from 34.1 ± 12 Gy to 29.5 ± 12 Gy. PTV coverage for re-optimized plans is comparable with original plans. PTV_7000 V65 Gy[%], D95%[Gy] and V80 Gy[%] for original plans are 98.7 ± 2.7%, 70 ± 2.3 Gy, 1.2 ± 4.5 Gy; and those for the re-optimized plans are 99.9 ± 0.1%, 70.8 ± 0.6 Gy, and 0.05 ± 0.1%, respectively. Knowledge-engineering based models predicted IMRT plans of better target coverage and OAR sparing as compared with submitted 0522 cases. All of IMRT plans with the worst score of deviation unacceptable from RTOG 0522 were improved to scores of variation acceptable or better with Knowledge based planning (KBP) re-optimization. These models will be used as part of the clinical trial quality assurance program to improve the quality of treatment plans submitted for clinical trials.